New-Employee Onboarding: A Task List That Connects HR, Equipment, Access, and Payroll
Back to blog
automation·September 6, 2026·4 min read·By Yehonatan Saadia

New-Employee Onboarding: A Task List That Connects HR, Equipment, Access, and Payroll

Build an onboarding dependency graph: one person record, real lead times, backward scheduling from the start date, and what actually counts as a completed task.

Key takeaways

  • One person record with one identifier is the root of everything. Four systems each holding their own spelling of the name is the reason the first payslip comes out wrong.
  • The schedule is built backwards from the start date using lead times. A laptop that takes seven business days to arrive will not be there if the order opens five days before.
  • Access is granted by a defined role, not by "the same as Dana". Copying a colleague's permissions copies everything they accumulated over years.
  • A completed task is represented by an artefact somebody can check: an equipment serial number, the account name that was created, the payroll employee id. A tick box is not an artefact.
  • Every grant has a matching revoke. The mirror image of onboarding is offboarding, and teams that do not build them together find active accounts belonging to people who left a year ago.

Onboarding is a dependency graph, not a flat list. It has one root - a person record with a single identifier - and one convergence point on day one, and every task in between is scheduled backwards according to its own lead time. This describes process and data flow, not obligations or entitlements.

One root: the person record

The first step is not opening an account or ordering equipment. It is creating one record that everything else points at. It is created the moment the signed offer comes back, and holds an internal identifier, the full name in one spelling, a start date, a role and a manager.

  • person_id is the identifier every other system stores. Not a name, not an email address, not a national id number.
  • The email address is derived from the record, never the reverse. Someone who changes their name later does not lose the link.
  • The start date is the field every task schedules from. Changing it must reschedule all of them, not merely raise an alert.
  • The ת"ז and bank details go to payroll and HR only, and are never copied into an IT ticket. Data minimisation is a design decision as much as a policy one.
  • Microsoft Entra implements exactly this pattern and names three event types - Joiner, Mover and Leaver - with tasks scheduled relative to the value of the employeeHireDate attribute.

Why does the laptop arrive late even when the checklist was completed?

Because a checklist describes what has to happen, not when it has to start. A task with a seven-business-day lead time must open seven business days ahead, and a flat list shows it on the same line as a task that takes ten minutes. The fix is backward scheduling: every task carries a lead_time, and the system computes start_by from the start date.

TaskOwnerDepends onLead timeScheduled for
Signed offer returnedHR--D-14
Create person recordHRsigned offerimmediateD-13
Order laptop and phoneProcurementperson record7 business daysD-12
Decide role and access bundleManagerperson record-D-7
Register with payrollPayrollperson record plus employee documents-D-5
Create directory account and mailboxITperson record1 business dayD-3
Attach role access bundleITaccount plus role decision1 business dayD-2
Desk, badge, parkingFacilitiesstart date2 daysD-2
Day-one schedule and buddyManager--D-4
First access reviewManagerall grants-D+30

The table also shows why the start date is a dangerous field. Moving it a week forward changes ten dates, and if the system only raises an alert about the change, seven of those tasks stay on the old date.

What counts as a completed onboarding task?

An artefact somebody can verify without asking anyone. The difference between "marked done" and "has an artefact" is the difference between an auditable process and a list of promises.

  • Equipment order: a serial number and an asset number recorded against person_id.
  • Account creation: the actual username, not the word "created".
  • Access: the name of the role granted and the systems it resolves to.
  • Payroll: the employee id returned by the payroll system.
  • Badge: the badge number and the door groups it opens.
  • Training: a completion date per module, not "started training".

Those artefacts are also what lets you answer one question on the morning of day one: what is missing. A list of ticked boxes cannot answer it. The customer-side equivalent, where the same pattern recurs with different artefacts, is described in automating customer onboarding.

Access: a defined role instead of a copy of a colleague

"Give him the same access as Dana" is the shortcut that produces the largest permissions problem in most organisations. Dana has been there four years, changed roles twice, and accumulated access she no longer needs. Copying duplicates all of it and creates a new joiner with a veteran's reach.

The correct pattern is one layer in the middle: a role. Instead of twenty direct grants there is one role mapped to a bundle of systems, and the grant is of the role. NIST SP 800-53 treats account management as its own control family covering creation, modification, periodic review and disabling. Three practical rules follow:

  1. Each role is defined once and documented: which systems, at what level, and who approves an exception to it.
  2. An exception is a separate grant with an expiry date, so temporary access does not quietly become permanent.
  3. A first review thirty days after the start date, where the manager confirms what is actually being used and removes the rest.

The other side of the same coin is leaving. Every grant in the bundle has a matching revoke, and it should be created at the moment the grant is created rather than reconstructed later. Training on the systems the new joiner now has access to is covered in CRM and ERP training and onboarding.

Where the process actually breaks

  • An offer signed two days before the start. Every lead time collapses and the process becomes a chain of phone calls.
  • Payroll registered without a linked identifier, so the person exists in HR and in payroll as two separate entities.
  • A role decision made on day one. IT waits, the new joiner sits with no access, and the workaround is an over-broad permission.
  • Equipment that arrived but was never recorded against person_id, so at offboarding nobody knows what to collect.
  • A start date changed in HR only, leaving the dependent tasks on the old date.
  • A part-time or second-site joiner run through the identical template, generating tasks that do not apply.

Sources

#employee onboarding#HR operations#access provisioning#equipment#workflow#Access

Frequently asked questions

How far ahead should the process start?

By the longest lead time in the graph, not by feel. If a laptop order takes seven business days and badge issuance takes two, the first task opens seven business days before the start date. The way to find that number is to measure actual lead times across three onboardings and update the fields.

Is a dedicated system necessary, or is a spreadsheet enough?

A spreadsheet works while dependencies are known and volume is low, but it does not schedule backwards and it does not alert. The signal that it is time to move is a start-date change: in a system that is one action moving ten tasks, and in a spreadsheet it is ten manual edits, one of which gets forgotten.

Who should own the process end to end?

One party holding the graph, usually HR, while each task belongs to the team that performs it. Owning the process is not doing the tasks; it is being accountable that every task is scheduled, assigned and reported with an artefact. Without central ownership each team sees only its own slice and nobody sees a missing step.

What do you do when someone starts before the equipment arrives?

Define a fallback path as part of the process rather than improvising: a loan machine from a pool, limited access for a period, and a task with a target date to swap it. What matters is that the fallback is recorded, otherwise the loan machine stays for two years and the asset register describes equipment that does not exist.

Keep reading

Related service

Business Automation

I build custom automations that remove repetitive work end to end.

Learn more →←

About the author

Yehonatan Saadia

Freelance automation, web & MVP developer

I'm Yehonatan Saadia, a senior developer who builds business automation, custom websites, and MVPs for small and mid-sized companies across the US, Europe, and Israel. These guides come from real client work, not theory.

Work with me

Have a project like this?

Tell me what you're trying to automate or build. I reply within 24 business hours with a few targeted questions, then we walk through it on a free 30-minute call, with no commitment. You come away with a scope, a timeline and a fixed price - or a straight answer that it isn't worth building.